结合运动想象和动作观察的伸-握动作识别方法

A recognition method for spread and grasp actions of a hand that combines motion imagination and action observation is proposed to obtain electroencephalogram (EEG) feature that is directly relevant to motor imagery of spread and grasp movements of the hand and to effectively identify fine actions. First, a compound paradigm that combines motor imagery and action observation (short for compound paradigm) is designed to enhance activation of the human cerebral cortex associated with the spread and grasp actions of a hand. Activation rules of the cerebral cortex are investigated under motor imagery, action observation and compound paradigm. Then, EEG is reconstructed for each channel by using a wavelet reconstruction and the multiscale entropy of EEG is calculated. EEG features of spread and grasp actions of the hand are extracted through a common spatial pattern. Finally, both the 'spread' and 'grasp' actions are identified from EEG features based on a linear discriminant method. Experimental results and comparisons with a single paradigm of motor imagery or action observation show that the proposed method effectively activates neurons in motor cortex and occipital cortex in low frequencies (5-20 Hz), and the multi-scale entropy of 'spread' action from reconstructed EEG is obviously larger than that of 'grasp' action. The fine intention perception of spread and grasp actions of one hand is realized, and an average accuracy rate of 85.7% is achieved.